Improving the Prediction of Potential Kinase Inhibitors with Feature Learning on Multisource Knowledge.

Area of Science:

• Computational drug discovery
• Bioinformatics
• Systems biology

Background:

• Kinase inhibitors are crucial for treating human diseases.
• Existing computational methods for kinase inhibitor identification often rely on limited features, such as sequence information.
• There is a need to leverage multisource knowledge for enhanced prediction of kinase inhibitors.

Purpose of the Study:

• To propose a novel computational framework, FLMTS, for improving kinase inhibitor prediction.
• To aggregate feature information from multisource knowledge for more effective prediction.
• To enhance the identification of potential kinase inhibitors in drug discovery.

Main Methods:

• FLMTS employs a random walk with restart (RWR) algorithm to integrate multiscale information within a heterogeneous network.
• Combined information from the network is utilized as features for both compounds and kinases.
• A random forest (RF) model is used to predict unknown compound-kinase interactions based on these aggregated features.

Main Results:

• FLMTS demonstrated significant performance improvements compared to existing state-of-the-art methods.
• Case studies validated the reliability and effectiveness of the FLMTS framework.
• Pathway enrichment analysis confirmed FLMTS's capability in accurately predicting disease-related signaling pathways.

Conclusions:

• The FLMTS computational framework effectively aggregates multisource knowledge for kinase inhibitor prediction.
• FLMTS achieves superior prediction performance over current state-of-the-art methods.
• This approach holds promise for advancing drug discovery and development.